170 likes | 192 Views
Explore advanced image compression codecs like MPEG-4 and H.264 for low-bitrate transmission over long-range sensor networks. Learn about DCT, quantization, and Huffman encoding methods. Discover ways to optimize energy consumption and improve image transmission in outdoor environments.
E N D
IMAGE COMPRESSION FOR WIRELESS OUTDOOR SENSOR NETWORKS Indrit Enesi, Elma Zanaj, Bexhet Kamo, Vladi Kolici, Olimpjon Shurdi Polytechnic University of Tirana, Albania
ISSUE • A NETWORK OF INTERCONNECTED DEVICES, CAPABLE OF RETRIEVING MULTIMEDIA CONTENT FROM THE ENVIRONMENT • NODES WITH LIMITED RESOURCES • POWER • BANDWITH • ENERGY
OBJECTIVE • A NEW IMAGE COMPRESSION CODEC • SUITABLE FOR WSN LOW-BITRATE IMAGE TRANSMISSION OVER LONG-RANGE OUTDOOR SENSOR NETWORKS
APPROACHES FOR LOW-BITRATE VIDEO CODING • WAVEFORM-BASED CODING • ON THE 2-D IMAGE PLANE • MODEL-BASED CODING • A MODEL OF THE THREE-DIMENSIONAL • THE 2-D IMAGES ARE ANALYZED AND SYNTHESIZED • VERY LOW BITRATE • VERY COMPLICATED TECHNIQUE • BLOCK-BASED HYBRID CODING • MPEG-4 [3] AND H.264 • ENCODER MUCH HIGHER COMPLEXITY THEN DECODER
PLATFORM • HW • FLECKTM OUTDOOR SENSOR NETWORKS • long-range radio • solar-charging circuitry • SW • TINYos
IMAGE COMPRESSINGPREPROCESSING • CCD 2D ARRAY SENSORS • YUV 4:2:0 • QVGA (320 X 240 pixels) • Packet 16 bits, 8 bit luminance, 8 bit chrominance
PREDICTION • CORRELATION BETWEEN PIXELS IN TEMPORAL DOMAIN • SKIP BLOCKS – ONLY INFORMATION • INTRA-BLOCK – ENCODING WITHOUT REFERENCE • INTER-BLOCK – ENCODING WITH REFERENCE FRAME
DCT • A MATRIX OF 8X8 FREQUENCY COEFFICIENTS PER BLOCK • 12 MULTIPLICATIONS PER 8-POINT TRANSFORM • MULTIPLICATIONS ARE DONE IN PARALLEL • MOST INFORMATION IN THE BLOCK WILL BE CONCENTRATED TO A FEW LOW-FREQUENCY COMPONENTS
ENTROPY CODING • COMBINATION OF RUN-LENGTH ENCODING (RLE) AND HUFFMAN ENCODING • PREDICT THE DC COEFFICIENT (EXCEPT THE FIRST BLOCK OF THE ROW) • EACH SYMBOL TWO PARTS • SYMBOL OF THE CATEGORY • ADDITIONAL BITS WITHIN A CATEGORY
DC COEFFICIENTS SYMBOL CATEGORIES FOR PREDICTED VALUES OF DC COEFFICIENTS SYMBOLS - HUFFMAN ENCODING
AC COEFFICIENTS • RLE AND HUFFMAN ENCODING • EACH HUFFMAN SYMBOL TWO PARTS • FIRST PART - RLE (NUMBER OF ZERO ELEMENTS) • SECOND PART – CATEGORY FOR NONZERO ELEMENT
HUFMMAN TABLES • DIFFERENT TABLES FOR AC AND DC COMPONENTS • DIFFERENT TABLES FOR LUMINANCE AND CHROMINANCE COEFFICIENTS • TOTAL FOUR TABLES
HUFFMAN TABLES FOR DC COEFFICIENTS DIFFERENCES LUMINANCE CHROMINANCE
RESULTS • A SYSTEM TO TRANSMIT COMPRESSED IMAGES OVER A WSN TO A DATABASE • THE COMPRESSION DEPENDS ON THE CONTENT, BETWEEN 90% AND 99% • MORE CAMERAS ADDED, SCHEDULATION, TO GET AND TRANSMIT IMAGES AT DIFFERENT TIMES. • SAVING IN ENERGY CONSUMPTION - POWER UP THE CAMERA, GET A SEQUENCE OF IMAGES, PARAMETER • IMAGES ARE RECONSTRUCTED TO A VIDEO CLIP AT THE BASE STATION.
FUTURE WORKS • POWER SAVING, VERY TIGHT SCHEDULING BETWEEN CAMERAS. • A MODIFICATION, NODES PUT TO SLEEP MOST OF THE TIME TO SAVE POWER. • ERROR CONTROL FOR IMAGES REQUIRES A CROSS-LAYER DESIGN BETWEEN THE CODING AND THE NETWORK • REDUCE THE PACKET HEADER OVERHEAD • TRANSPORT AND ROUTING LAYER, WHEN THE BURSTY VIDEO OR IMAGE DATA IS MIXED WITH THE OTHER SENSOR DATA. • DISTRIBUTED SOURCE CODING CERTAINLY IS INTERESTING